pub fn dice_loss(
graph: &mut Graph,
prediction: NodeId,
target: NodeId,
smooth: f32,
) -> Result<NodeId, ModelError>Expand description
Dice loss for segmentation.
DiceLoss = 1 - (2 * |P ∩ G| + smooth) / (|P| + |G| + smooth)
where prediction and target are probability maps.